when do startups that exploit patented academic knowledge ...€¦ · when do startups that exploit...
TRANSCRIPT
When Do Startups That Exploit Patented Academic Knowledge Survive?1
Atul Nerkar2 Columbia University
Scott Shane University of Maryland
Abstract
Researchers have generally suggested that new technology firms should exploit radical technologies with broad scope patents to compete with established firms, implying that new firms founded to exploit university inventions will be more likely to survive in all industries if they possess these attributes. However, the existing empirical evidence indicates that the effectiveness of these two dimensions of new firm strategy is contingent on the industry environment, specifically industry concentration. In this paper, we explain why this industry-specific relationship should exist and use a unique data set of new technology ventures originating at Massachusetts Institute of Technology to test our arguments. Key Words: Entrepreneurship (M13); Management of Technological Innovation and R&D (O32).
1 Authors are listed alphabetically. Submitted to the Special Issue on “Economics of Intellectual Property at
Universities” of the International Journal of Industrial Organization. We thank Ajay Agrawal and John Scott for
helpful comments on an earlier draft of this paper.
2 Corresponding author: Graduate School of Business, 721 Uris Hall, 3022 Broadway, New York, NY 10027; Tel:
(212) 854 4431; Fax: (212) 316 9355; Email: [email protected]
2
Universities have traditionally been considered an important source of new technology
(Barker, 1985; Jaffe, 1989; Rosenberg and Nelson, 1994). For example, the Internet originated
as an electronic discussion group for a community of physicists in Switzerland. Similarly, the
ideas of nuclear fission and fusion were first discussed at Columbia University. Moreover, in
recent years, the role of universities in commercial technology development has become even
more important as partnerships with the private sector (Siegel et al, 2002; Siegel et al
forthcoming, Hall et al, 2002) and technology licensing (Thursby and Thursby, 2002), have
grown.
One of the most important developments in university technology commercialization in
recent years has been the significant rise in the creation of new companies as vehicles to exploit
university inventions (DiGregorio and Shane, 2003; Feldman et al, forthcoming). In fact, many
university technologies, from Genentech’s use of recombinant DNA to Lycos’ Internet search
engine, have led to the creation of new technology companies (Zucker, Darby and Brewer,
1998).
Despite the high profile of a few of the most successful of these university startups, many
of them have not been very successful. This article seeks to explain why some new companies
founded to commercialize patented university inventions survive when others do not. While
many explanations have been offered for the differential survival of university start-ups,
including the psychology of the founders (Roberts, 1991), their social ties (Shane and Stuart,
2002), or the university from which they come (Saxenian, 1990), no researchers have examined
the effects of the technology exploited by the start-up on its survival.
Several researchers have found that the nature of a firm’s technology base influences its
survival (Lerner, 1994; Tushman and Anderson, 1986; Henderson, 1993; Utterback, 1994;
3
Christiansen and Bower, 1996). However, these researchers have focused on explaining why
established firms fail in the face of technological innovation, rather than on explaining why new
firms survive entry. For example, Henderson (1993), in a study of the photolithographic
industry, found that entrants are more successful than incumbents at exploiting radical new
technology, while Tushman and Anderson (1986) found, across a range of industries, that
competence-destroying new technologies tend to be exploited by new companies. In their study
of the disk drive industry, Christiansen and Bower (1996) found that established companies tend
to be replaced by new companies when a new technology was not valuable to their mainstream
customers. Nevertheless, it is entirely possible that the nature of firm’s technology base explains
why established companies fail, without offering any explanation for why new companies might
survive entry.
While the studies mentioned above do not directly test the factors that influence the
survival of new technology firms, they do suggest the following argument: New technology
firms are likely to survive if they exploit radical technologies that cannot be imitated in the
founding period when a firm’s marketing and manufacturing assets are being established,
thereby allowing the new firm to undermine the advantages that established firms have in
pursuing incremental technologies (Merges and Nelson, 1990; Teece, 1986). Therefore, new
firms founded to exploit university inventions should be more likely to survive if they exploit
radical technologies with broad scope patents (Shane, 2001).
However, even casual observation of the empirical evidence on the survival of new
technology companies suggests that the benefits of these attributes appear to be quite industry-
specific. While the exploitation of radical inventions with broad scope patents appears to allow
new firms to survive competition with established firms in some industries (Christiansen and
4
Bower, 1996; Foster, 1986; Romanelli. 1989), this strategy appears less effective in others (Gans
and Stern, 2000; Teece, 1986).
We believe that industry differences in the effectiveness of a new firm strategy to exploit
radical technology with broad scope patents can be explained by considering the nature of
industry concentration. The survival of a new technology company requires the creation of the
marketing and manufacturing assets necessary to exploit the technological opportunity, and the
possession of a valuable technology that undermines the advantages of established firms and that
can be protected against immediate imitation by others (Gans and Stern, 2003; Venkataraman,
1997). Holding the possession of a radical technology and broad scope patent protection
constant, the concentration of an industry inhibits the survival of new firms by making it more
difficult for their founders to create the marketing and manufacturing assets necessary to exploit
the technological opportunity. As a result, even if the new firm possesses a radical technology
that undermines the competence of established firms and broad scope patents that protect the
technology, in concentrated industries, firm founders are often unable to create the set of assets
necessary to serve customers in a way that allows the firm to survive (Teece, 1986).
We provide support for these arguments by conducting an empirical test of the survival of
128 new technology companies founded between 1980 and 1996 to exploit MIT-assigned
inventions. Our results show that new firm survival is enhanced by radical technology and broad
scope patent protection only in fragmented industries.
Radicalness of Technology and Industry Concentration
The standard explanation for the effectiveness of a new firm strategy to exploit new
technology is that the firm will be successful if it exploits a radical technology. Radical
5
technology undermines the advantages that established firms have in making incremental
improvements to technology, undermines firm competence, and turns existing customer
relationships into liabilities rather than assets (Utterback, 1994; Christiansen and Bower, 1996;
Tushman and Anderson, 1986). However, this explanation is incomplete, as it does not consider
the nature of competition from established firms in the product market. Such competition will
determine whether a firm formed for exploiting radical technology will survive entry because the
introduction of radical technology by new firms may spur actions by established firms that
inhibit the survival of new firms.
To survive, new firms must build an organization and acquire assets that will be used in
conjunction with their radical technology. This process is more difficult in concentrated
industries than in fragmented industries. First, in concentrated industries, the marketing and
manufacturing assets necessary to exploit a technology lie in the hands of a few established
firms, which tend to acquire ownership of these assets to mitigate contracting problems
(Williamson, 1985; Teece, 1986). Because new firms do not have these assets in place at the
time of founding, they need to build them. When the needed assets are controlled by a few large,
existing firms, there are fewer parties for the new firms to work with, thereby increasing the
difficulty of establishing an agreement with one of them to obtain needed assets. In contrast, in
fragmented industries, the assets necessary to exploit the new technology are available from a
wide number of industry players, minimizing bargaining problems in efforts to obtain access to
these assets (Williamson, 1975).
Second, the average size of firms is larger in more concentrated industries, requiring the
amount of marketing and manufacturing assets that new firms have to build to be larger in
concentrated industries than in fragmented industries (Mansfield, 1981; Kamien and Schwartz,
6
1975). Consequently, the scale of cooperation necessary with established players is larger in
more concentrated industries. Given capital market imperfections, new firms find it difficult to
build up these assets on a scale that is cost effective with established players (Audretsch and
Mahmood, 1991; Geroski, 1995), undermining their ability to survive.
Third, in concentrated industries, established firms have cost advantages and market
power, which allows them to drive out the new competitor (Galbraith, 1956; Kamien and
Schwartz, 1975). Not only can they jointly work to deter entry by others (Acs and Audretsch,
1989), large established firms can price their products at a level that makes entry unprofitable for
the new firm, hindering its survival.
Fourth, in fragmented industries, new firm entry does not necessarily impact the efforts
of market leaders to serve their customers because the target customers of the new firm can
belong to small, established players (Eisenhardt and Schoonhoven, 1990). However, in
concentrated industries, the establishment of the new firm impacts the efforts of market leaders
to serve their customers because the target customers of new firms belong to large established
players. As a result, entry by the new firm is more likely to invoke retaliation by a large
established firm (Acs and Audretsch, 1987; Romanelli, 1989). Because the large established
firm has the power to compete against the new firm through the exploitation of lower costs or
superior sales efforts, the new firm is unable to gain a foothold in the market unchallenged by
those who can drive it out of business. As a result, the new firm finds it hard to enter the market
successfully and its survival is impaired. Thus,
Hypothesis 1: The relationship between radicalness of technology and the survival of a new venture is moderated by concentration within the industry, i.e. radicalness increases the likelihood of firm survival more in fragmented markets than in concentrated markets
7
Scope of IP Protection and Industry Concentration
New technology firms begin without any competitive advantages other than that
embedded in their new technology itself. Yet, as we indicated above, to survive, new firms must
develop manufacturing and marketing assets that are used in conjunction with their new
technology. Therefore, to survive, the new firm uses intellectual property protection to defend
the new technology against imitation until such time as its marketing and manufacturing assets
can be put into place (Teece, 1986). Broad scope patents facilitate this transition because they
provide better protection than narrow scope patents. As Merges and Nelson (1990:839) explain,
“the broader the scope, the larger number of competing products and processes that will infringe
the patent.”
The use of broad scope patents to defend the new firm against imitation by established
firms until the marketing and manufacturing assets can be put in place works well in fragmented
markets where it is relatively easy to gain access to marketing and manufacturing assets without
challenging market leaders. However, this strategy works poorly in concentrated industries.
First, when the industry is concentrated, the scale and scope of operations necessary to compete
in the industry are so large that it is not possible for the new firm to raise money from the capital
markets and create necessary assets before competitors find ways around the new firm’s
intellectual property protection (Audretsch and Mahmood, 1991). Second, when the industry is
concentrated, the large established players could compete with the new technology firm
effectively even if the new firm has effective intellectual property protection (Acs and
Audretsch, 1987). Leading firms in concentrated markets have cost and market power
advantages that allow them a basis of competition that offsets the advantages provided by the
8
new firm’s patented technology (Acs and Audretsch, 1989; Galbraith, 1956). Third, when the
industry is concentrated, any effort by the new firm to introduce its new technology requires it to
target the large established firm’s customers. This effort to take customers from the large
established firm provokes retaliation from better funded and more connected established firm
that invests money and time in undermining the new firm’s patent protection through lawsuits
and other tactics (Eisenhardt and Schoonhoven, 1990; Romanelli, 1989). Fourth, when the
industry is concentrated, the new firm needs to obtain more assets from the few established firms
that dominate the market. These firms are reluctant to work with the new firm as its technology
might undermine their market positions. Therefore, even though the new firm has protected
intellectual property, it cannot obtain the assets that it needs to compete in the industry (Teece,
1986). As a result, in concentrated industries, new firms need more than effective intellectual
property protection to survive and broad patent scope is not a sufficient condition for survival.3
Hence:
3 Readers will note that our arguments about the effect of concentration on the survival of new technology
companies do not discuss the incentives of incumbent firms to innovate, even though this is a central focus of the
economics of innovation (Baldwin and Scott, 1987). Although a large theoretical literature has debated whether
monopolists have a strong (Gilbert and Newbury, 1982) or weak (Reinganum, 1983) incentives to innovate, we do
not believe that models of the incentives of established firms to innovate focus on the key factors that influence the
survival of university spinoffs in concentrated industries. First, the theoretical literature largely assumes that the key
behavioral driver behind industry concentration is its effect on established firms’ decisions about whether or not to
innovate. However, most of the university spinoffs that we study were founded after established firms, which had
invested in the research that led to the new university technologies, had decided not to pursue those technologies
themselves. We believe that it is a stretch to argue that the main effect of industry concentration on competition
between established firms and startups is that established firms have greater incentives to innovate when we observe
that many established firms from concentrated industries invest in university research and then pass on innovating
9
Hypothesis 2: The relationship between the scope of IP protection and the survival of a new venture is moderated by industry concentration, i.e. patent scope increases the likelihood of firm survival more in fragmented markets than in concentrated markets.
RESEARCH METHODS
The data used for this paper was collected from the Technology Licensing Office (TLO)
of the Massachusetts Institute of Technology. Like many other research universities, MIT often
patents commercially useful inventions that are developed by their faculty, staff or students and
that emerge from work making material use of MIT resources. The TLO’s objective is to
commercialize MIT technology. Both established and startup companies license MIT
the technologies that emerge from research. Second, we believe that, if the effect of concentration on the survival of
startups were through its effect on established firms’ incentive to innovate, we would see established firms adopt
other behavior consistent with that process. As Gilbert and Newbury (1982) argue, a monopolist has more incentive
to innovate than a new entrant because of the threat of entry to its monopoly rent. This argument would suggest that
established firms adopt behaviors consistent with deterring startups from developing technologies that could
potentially undermine the established firms’ monopoly rents. However, we do not observe established companies
engaging in strategic behaviors consistent with this approach. For instance, we fail to observe established
companies in concentrated industries licensing university inventions and then shelving them, as a way to prevent
new entrants from exploiting the new technology, even though established firms have a right of first refusal to
license inventions that come from research that they fund at universities like MIT. Third, our empirical
investigation focuses on the survival of new technology firms, and does not include information about the behavior
of established firms. We believe that it is better not to assume that the survival of new firms depends on the
innovative behavior of established firms when we have no information about the effect of the start-up firms on the
innovative behavior of those firms. Our arguments about the effect of concentration on the survival of new firms
would hold even if industry concentration has no effect (either positively or negatively) on the incentives of
established firms to innovate.
10
technology. Given our research question, we focused only on new ventures and exclude all
licensing activity that involved established firms. We gathered information on 128 of the 134
firms that were founded to exploit inventions made at MIT between 1980 and 1996.
The TLO archives describe the contracts between MIT and its licensees, characteristics of
the licensed intellectual property, and the startups’ business plans. In addition, the TLO tracked
the sales growth and survival of its licensees. To corroborate, as well as to supplement, the TLO
data, we conducted unstructured interviews with company founders and consulted online
databases including Lexis/Nexis, Dialog Business Connection and ABI Inform. Finally, we
obtained additional information on the venture capital financing of these firms and the industry
in general from the Venture Economics and Venture One databases. The objective of the data
collection effort was to create detailed profiles of all firms from the time they were founded until
the time we stopped our observation or the time of firm failure.
Although MIT-based start-ups clearly do not constitute a representative sample of all
technology companies, these data have two advantages relative to other samples of startup firms.
First, our approach avoids excluding the large number of firms that fail at very young ages
(before they are recorded in directories), in large part because they are not able to secure external
financing. Therefore, there is no sample selection bias with respect to very young firms that do
not show up in industry directories. A second advantage of the TLO data is that there are no left-
censored observations, i.e. firms are observed from the time of founding. This allows us to
analyze firm survival using event history analytical techniques with reliable parameter estimates.
11
Method
The event we model in this paper is the failure of a startup founded to exploit MIT
intellectual property. We define a failure as a firm that ceases operations. We treat those
companies that enter markets and are acquired by established firms as successes because many
new companies established to exploit new technology aspire to acquisition as an exit strategy.
We analyze this transition from a going concern to a failed firm, which we define as a
bankruptcy or cessation of operations, in terms of the instantaneous transition rate (otherwise
known as the “hazard rate”), r, defined as
rk(t)= lim Pr(t<T< (t+ ∆t), D=k| T>∆t) (1)
∆t 0 ∆t
where k represents a failure. The variable T measures the time spent at risk of failure and the
probability Pr refers to the likelihood of experiencing a failure during the small interval from t to
∆ t, conditional on a firm having survived as of time t (Tuma and Hannan, 1984). The waiting
time clock in the firm event histories turns on at the time of founding. The transition of a startup
from a going concern to a failure is modeled using the approach discussed in Kalbleisch and
Prentice (1980). We created annual spells to update the values of the time changing covariates.
Each spell ends in censoring unless an event occurs within the focal firm year observation. We
used the Weibull model as it has a simple survivor function that is easy to manipulate and many
firm failure studies suggest that failure follows a Weibull distribution. The form that this model
takes is:
rk(t) =exp[( -βXi)1/ρ] where (2)
rk(t) is the instantaneous hazard function of failure while Xi is the vector of covariates and ρ is
the Weibull parameter.
12
Predictor Covariates
Technological Radicalness: Following Rosenkopf and Nerkar (2001), we measure
technological radicalness as the count of the number of United States Patent and Trademark
Office (USPTO) patent classes cited in prior art of the licensed invention outside of the patent’s
own class. Past research has shown that this measure captures the degree to which a technology
is radical or competence changing in a technical sense (Rosenkopf and Nerkar, 2001; Shane,
2001).
Patent Scope: We measure patent scope as a count of the number of International Patent
Classification (IPC) classes that a patent is classified into by the USPTO (Lerner, 1994). Because
the international patent classification is nested, the count of classes indicates a scale of patent
scope (Shane, 2001). Past research has shown that patents with a greater number of IPC classes
have more claims, face a greater likelihood of infringement and are more likely to be litigated
(Lerner, 1994).
Industry Concentration: We include a measure nature of competition as the C4 ratio, i.e.
the market share accounted by the top 4 companies in the 4-digit SIC code where the new
company is classified. We examine only the C4 concentration ratio in our analyses because
previous researchers have not found significant qualitative or statistical differences between the
C4 concentration ratio and a Herfindahl index (see, for example, Scott, 1993). Using data from
the Census Bureau, we include the concentration ratio in the year that the firm was founded.
Because industry concentration changes slowly over time, we do not expect that a one-time
measure of industry concentration will lead to errors in measurement. We use the categorization
made by MIT’s technology licensing office to assign firms to industries.
13
Control Covariates
Our focus in this paper is on technological entrepreneurship, a sector that is heavily
influenced by a host of factors at the environment, firm, and individual levels of analysis. Hence
we include a variety of control covariates:
Technical Fields: We control for inventions in the chemical or electrical fields with
dummy variables for these fields because the process and rate of new firm development differs
across types of technology. By controlling for these technical fields, we can partial out this type
of variation from the data.
Biotechnology Firms. We use a dummy variable of 1 to control for biotechnology firms
because biotechnology spin-offs from universities differ from other university spin-offs in two
fundamental ways. First, biotechnology is the only industry in which the locus of technology
creation lies in universities, suggesting that the survival patterns of university biotechnology
firms might be different from that of other university start-ups. Second, biotechnology firms
tend to require far more capital to commercialize their technologies than other university spin-
offs, and often go public without having products on the market. As a result, the survival
patterns for biotechnology firms might be systematically different from the survival patterns of
other university spin-offs.
Number of Firms in the Industry: Our measure of concentration is an indication of the
power wielded by the top four firms in the industry. We also want to control for the total number
of players in the industry to capture the number of firms competing in the industry. Using data
from the Census Bureau, we include a count measure of the firms at 4-digit SIC level in the year
14
that the firm was founded. We use the categorization made by MIT’s technology licensing office
to assign firms to industries. The variable is skewed and we use a logarithmic transformation.
Venture Capital in the Industry: Survival of firms is also linked to the availability of
resources in the environment. Using data from Securities Data Corporation’s venture capital
database, we include a variable that counts the total number of firms funded by venture
capitalists in the industry in the year that the firm was founded. We use the categorization made
by MIT’s technology licensing office to assign firms to industries. The variable is skewed and
we use a logarithmic transformation.
General Purpose. General-purpose technologies should enhance the survival of new
firms because these technologies can be used in a variety of applications. As a result, they
provide new firms with flexibility that is useful to overcoming technical and market risk (Shane,
2000). We include a dummy variable that takes the value of 1 if MIT records showed that the
technology is a general-purpose technology that could be applied in multiple fields.
Entrepreneur Is Inventor: Agency problems are less likely to cause the failure of a
venture if the entrepreneur is also the inventor (Holmstrom, 1989). We include a dummy
variable that takes the value of 1 if the person starting the firm is also the inventor.
Prior Knowledge of Problem Solved: Prior experience with technical problems can
reduce the likelihood of failure at solving them. In the case of technology ventures, founder
experience with the specific technical problem to be solved can reduce the likelihood venture
failure. We include a dummy variable that takes a value of 1 if the entrepreneur has prior
knowledge of the technical problem that the venture seeks to solve (Shane, 2000).
Inventor is Tenured at MIT: An invention patented by a senior faculty member may
influence the likelihood of failure of a firm started to exploit it because tenure might increase the
15
risks that the founder is willing to take or his ability to leverage his reputation to obtain
resources. We include a dummy variable that takes a value of 1 if the inventor is tenured at MIT
and 0 if he or she is not tenured.
Product Invention: Product inventions are more likely to fail than process inventions
because they need to be accepted by end customers and not just the founding firm (Utterback,
1994). Hence we include a dummy variable that takes a value of 1 if the licensed invention is a
product invention.
Exclusive License: New firms are more likely to succeed if they have exclusive access to
the intellectual property they are exploiting. Therefore, we also coded a dummy variable,
exclusive, which takes the value of 1 if the new venture has an exclusive right to use MIT
technology in a particular field of use.
Experience of Founding Team: We include variables that measure two different types of
founder business experience. The first measure (start up experience) counts the number of firms
started by the founding team in the past while the second measure (market experience) counts the
number of years of industry experience on the founding team. Both measures of experience
should reduce the likelihood of failure as prior research has shown that these dimensions of
human capital are important to new venture performance (Bates, 1990).
RESULTS
The descriptive statistics are reported in table 1, while table 2 reports the results from the
event history analysis predicting firm survival.
_______________________
Insert Tables 1and 2 here _______________________
16
Models 1 and 2 in Table 2 predict the hazard of firm failure as a function of
concentration and technological radicalness and patent scope respectively. These models indicate
that overall new technology firms are less likely to fail if they exploit radical technology and
have broad scope patents. Model 3 includes the interaction term between technological
radicalness and concentration, while Model 4 focuses on the interaction between patent scope
and concentration. (Due to problems of multicolinearity, we are unable to separate the effects of
the two interactions simultaneously.) The results offer support for our two hypotheses that
technological radicalness and patent scope reduce new firm failure overall, in the sense of what
the effect is without the interactive control for seller concentration. However, technological
radicalness and patent scope increase new firm failure in concentrated markets.
We also show the results for radicalness and concentration graphically in Figures 1 and 2.
These figures demonstrate the interaction effect between concentration and both radicalness and
patent scope on firm failure.
_________________________________
Insert Figures 1 and 2 here _________________________________
Our findings might be explained alternatively if industry concentration proxied for
unobserved industry characteristics that would also predict the interaction effects we observe.
We seek to rule out these alternative explanations. Tables 3 through 6 report a series of analyses
that we conducted to examine alternate explanations that industry concentration proxies for the
presence of dominant designs, industry R&D intensity, market growth rates and technological
maturity, and that the interaction between radicalness and these factors actually explain the
results that we showed in Table 2.
17
The first alternative explanation is that concentration is proxying for the presence of a
dominant design in certain industries because industry concentration is higher in industries that
have achieved a dominant design. One might expect that the introduction of a radical technology
would not be survival enhancing for a new firm if a dominant design existed in the industry. In
Table 3 we include a variable from the Yale Survey on Innovation (Levin et al, 1987) that
measures if there was a dominant design in the industry. The results shown in Table 3 indicate
that even when the interaction of radicalness and dominant design is measured, we still observe
the predicted relationship between radicalness and concentration.
The second alternative explanation is that industry concentration is proxying for the level
of R&D spending in an industry because more concentrated industries have more R&D
spending. One might expect that the introduction of a radical technology would not be survival
enhancing for a new firm if established firms had a high level of absorptive capacity or ability to
innovate. In Table 4 we include a variable that measures industry R&D intensity and its
interaction with radicalness. The results in Table 4 indicate that even when the interaction of
radicalness and R&D intensity is measured, we still observe the predicted relationship between
radicalness and concentration.
The third alternative explanation is that industry concentration is proxying a lack of
market growth because more concentrated industries tend to grow more slowly than less
concentrated industries. One might expect that the introduction of a radical technology would
not be survival enhancing if the market is growing slowly. Market growth and its interaction
with radicalness are included in Table 5. The results in Table 5 indicate that even when the
interaction of market growth and radicalness is measured, we still observe the predicted
relationship between radicalness and concentration.
18
The fourth alternative explanation is that industry concentration is proxying the age of the
technology because industries with older technologies tend to be more concentrated than
industries with younger technologies. One might expect that the introduction of a radical
technology would not be survival enhancing if the industry were old. We include the interaction
between age of the technology and radicalness in Table 6. The results in Table 6 indicate that
even when the interaction of age of the technology and radicalness is measured, we still observe
the predicted relationship between radicalness and concentration.
A fifth alternative explanation for our results is that they are an artifact of the Weibull
model assumption that the hazard of an event is a smooth function of time. Therefore in
unreported regressions, we reanalyze or results with a piecewise exponential model of the form:
rk(t) =exp[ γp + βXi]
where γp includes two duration period effects, Xi contains independent variables (some of
which vary over time), and β represents the parameters to be estimated. The piecewise
specification of duration dependence permits the rate to vary flexibly with duration (in this case
firm age) without requiring strong parametric assumptions. The age pieces we include are less
than four years and more than four years old respectively; the baseline rate is assumed to be
constant within each period, but is constrained across periods. The results are qualitatively the
same when we use the piecewise exponential model as when we use the Weibull model,
suggesting that the results are not explained by the assumptions of the statistical model.
Another challenge to our results is that strategies that lead to a higher probability of
failure might also lead to a higher expected payoff. As a result, having radical technology and
broad scope patents in fragmented industries might not be beneficial to startups because such a
technology base would reduce the likelihood of successful outcomes, such as achieving an initial
19
public offering. To rule out this alternative interpretation, we used the same regression model
shown in Table 2 to predict the hazard of the new firm achieving either an initial public offering
or being acquired by another firm. In unreported regressions, we found that the interaction
between concentration and both patent scope and technological radicalness has no statistically
significant effect on the hazard of achieving a positive outcome. Thus, we believe that having a
radical technology and broad scope patents in a fragmented industry reduces the failure of
university spin-offs without influencing their likelihood of achieving a positive outcome.
DISCUSSION
This article examined the interaction between radicalness of technology and patent scope
and industry concentration on the likelihood of firm failure for university start-up companies.
We examine a unique data set of 128 firms founded to commercialize technologies licensed from
MIT between 1980 and 1996 and show that technological radicalness and patent scope reduce
new firm failure only in the context of fragmented markets.
Limitations
This study is not without limitations. We measure radicalness as the number of patent
classes cited outside of a patent’s own patent class. While many scholars have employed this
measure of radicalness (e.g., Rosenkopf and Nerkar, 2001; Shane, 2001), researchers might ask
whether this construct should be measured differently. For instance, should these measures be
normalized by industry? Readers should note that our results are contingent on the construct
validity of the radicalness measure.
20
In addition, we measure patent scope as the number of international patent classes
assigned to the patent, consistent with the prior work of Lerner (1994) and Shane (2001).
Despite the evidence provided by Lerner (1994) of the construct validity of this measure,
researchers might ask if this measure should be normalized by industry. Given this question, we
must caution readers that the validity of our results is contingent on the construct validity of the
patent scope measure.
A second important limitation concerns the generalizability of our results. Our sample
consists of firms that were formed to exploit the intellectual property created by the
Massachusetts Institute of Technology. This institution is one of the largest generators of
university spin-offs in the United States. Given that some universities are more likely than
others to generate spin-offs to exploit their intellectual property (DiGregorio and Shane, 2003),
the Massachusetts Institute of Technology might not represent the general population of
universities. As a result, the patterns that explain the performance of new companies founded to
exploit the Institute’s intellectual property might not generalize to new companies founded to
exploit the intellectual property of other institutions. While we have no a priori reason to suspect
that our results are not generalizable, we also have no evidence to support their generalizability.
Therefore, readers should interpret our results with caution.
Implications
Our results indicate that two dimensions of the strategy of new technology companies
founded to exploit university inventions are industry-specific. Previous research has argued that
new technology firms will be more likely to survive if they exploit radical technology because
the advantages of established companies in exploiting incremental technology require new
21
competitors to exploit competence destroying new technology (Tushman and Anderson, 1986).
Our results provide insight into the empirical puzzle engendered by this argument. Why does the
exploitation of radical technology only appear to help new firms survive in certain industries?
We offer as explanation that a strategy to exploit competence-destroying radical technology as a
way for a new firm to compete only works in fragmented industries. In concentrated industries,
the exploitation of a radical technology fails to provide an advantage to new companies.
Concentrated industry environments hinder efforts of the new firm to build the manufacturing
and marketing assets necessary to compete.
Previous research has also argued that new technology firms will perform better if they
have broad scope patents because strong intellectual property protection is necessary to protect
their technology from imitation while they create the marketing and manufacturing assets
necessary to exploit their technologies (Merges and Nelson, 1990). We show that this strategy is
also contingent on the entrepreneur founding a company in a fragmented industry. In a
concentrated industry, the difficulty of creating these assets makes this strategy problematic.
In sum, we believe that this paper opens a new avenue of inquiry into the ways in which
the industry environment in which new university technology start-ups compete interacts with
their strategy to influence their survival. While our study only offers insight into one small way
in which strategy-environment interaction influences new technology firm survival, we hope that
it will spur other researchers to examine this question, rather than simply assuming that a given
strategy is effective across all industry environments.
22
REFERENCES Acs, Z., and D. Audretsch, 1987, Innovation, market structure, and firm size. The Review of
Economics and Statistics 71, 567-574. Acs, Z., and D. Audretsch, 1989, Small firm entry in US manufacturing, Economica 56, 255-
266. Audretsch, D., and T. Mahmood, 1991, The hazard rate of new establishments. Economic Letters
36, 409-412. Baldwin, W., and J. Scott, 1987. Market structure and technological change, (Reading, UK:
Harwood Academic Publishers). Barker, R., 1985, Bringing science into industry from universities. Research Management, 28(6),
22-25 Bates, T., 1990, Entrepreneur human capital inputs and small business longevity, The Review of
Economics and Statistics 72(4), 551-559. Christensen C., and J. Bower, 1996, Customer power, strategic investment, and the failure of
leading firms, Strategic Management Journal 17, 197-218. DiGregorio, D., and S. Shane, 2003, Why do some universities generate more startups than
others? Research Policy 32(2), 209-227. Eisenhardt, K., and K. Schoonhoven, 1990, Organizational growth: Linking founding team,
strategy, environment, and growth among U.S. semiconductor ventures, 1978-1988, Administrative Science Quarterly 35, 504-529.
Feldman, M., I. Feller, J. Bercovitz, R. Burton, Forthcoming, Equity and the technology transfer
strategies of American universities, Management Science. Foster, R., 1986, Innovation: The attackers advantage. (Summit Books: New York). Galbraith, J., 1956, American Capitalism, (Houghton Mifflin: Boston). Gans J., and S. Stern, 2000, Incumbency and R&D incentives: Licensing the gale of creative
destruction, Journal of Economics and Management Strategy 9, 485-511. Gans, J., and S. Stern, 2003, The Product Market and the Market for Ideas: Commercialization
Strategies for Technology Entrepreneurs, Research Policy 32(2), 333-350. Geroski P., 1995, What do we know about entry? International Journal of Industrial
Organization, 421-440.
23
Gilbert, R., and Newbury, 1982, Preemptive patenting and the persistence of monopoly,
American Economic Review 75, 514-526. Hall, B., A. Link, and J. Scott, Forthcoming, Universities as research partners. Review of
Economics and Statistics. Henderson, R., 1993, Underinvestment and incompetence as responses to radical innovation:
Evidence from the photolithographic alignment equipment industry, The Rand Journal of Economics 24(2), 248-271.
Holmstrom B., 1989, Agency costs and innovation, Journal of Economic Behavior and
Organization 12(3), 305-327. Jaffe A., 1989, Real effects of academic research, American Economic Review, 957-970. Kalbleisch, J., and R. Prentice, 1980, The Statistical Analysis of Failure Time Data, (Wiley: New
York). Kamien, M., and N. Schwartz, 1975, Market structure and innovation: A survey, Journal of
Economic Literature 13, 1-37. Lerner, J., 1994, The importance of patent scope: An empirical analysis, RAND Journal of
Economics 25(2), 319-333. Levin, R., A. Klevorick, R. Nelson, and S. Winter, 1987, Appropriating the returns from
industrial research and development, Brookings Papers on Economic Activity 3, 783-832. Mansfield, E., 1981, Composition of R&D expenditures: Relationship to size of firm,
concentration, and innovative output, Review of Economics and Statistics 63, 610-615. Merges, R., and R. Nelson, 1990, On the complex economics of patent scope, Columbia Law
Review 90(4), 839-916. Reinganum, J., 1983, Uncertain innovation and the persistence of monopoly, American
Economic Review 73, 741-748. Roberts, E., 1991, Entrepreneurs in High Technology, (Oxford University Press: New York). Romanelli E., 1989, Environments and strategies of organization start-up: Effects on early
survival, Administrative Science Quarterly, 369-387. Rosenberg, N., and R. Nelson, 1994, American universities and technical advance in industry.
Research Policy, 323-348.
24
Rosenkopf, L., and A. Nerkar, 2001, Beyond local search: Boundary-spanning, exploration, and impact in the optical disk industry, Strategic Management Journal 22, 287-306.
Saxenian, A., 1990, Regional networks and the resurgence of Silicon Valley. California
Management Review 33(1), 89-113. Scott, J., 1993, Purposive Diversification and Economic Performance, (Cambridge University
Press: Cambridge, UK). Shane, S., 2000, Prior knowledge and the discovery of entrepreneurial opportunities,
Organization Science 11, 448-469. Shane, S., 2001, Technology opportunity and firm formation, Management Science 47(2), 205-
220. Shane S, and T. Stuart, 2002, Organizational endowments and the performance of university
start-ups, Management Science 48(1), 154-170. Siegel, D., J. Poyago-Theotoky, and J. Beath, 2002, Universities and fundamental research:
Reflections on the growth of university-industry partnerships, Oxford Review of Economic Policy 18(1), 10-21.
Siegel, D., D. Waldman, L. Atwater, and A. Link. Forthcoming. Commercial knowledge
transfers from universities to firms: Improving the effectiveness of university-industry collaboration, Journal of High Technology Management Research.
Teece D., 1986, Profiting from technological innovation: Implications for integration,
collaboration, licensing and public policy, Research Policy, 285-306. Thursby, J., and M. Thursby, 2002, Who is selling the ivory tower? Sources of growth in
university licensing, Management Science 48, 90-104. Tuma, N., and M. Hannan, 1984, Social dynamics: models, and methods, (Academic Press, New
York). Tushman, M., and P. Anderson, 1986, Technological discontinuities and organizational
environments, Administrative Science Quarterly, 439-465. Utterback, J., 1994, Mastering the Dynamics of Innovation, (Harvard Business School Press:
Boston). Venkataraman, S., 1997, The distinctive domain of entrepreneurship research: An editor's
perspective, In J. Katz and R. Brockhaus (Eds.), Advances in Entrepreneurship, Firm Emergence, and Growth 3, (JAI Press: Greenwich, CT), 119-138.
25
Williamson, O., 1975, Markets and Hierachies: Analysis and Antitrust Implications, (Free Press: New York).
Williamson, O., 1985, The Economic Institutions of Capitalism, (Free Press: New York). Zucker L, M. Darby, M. Brewer, 1998, Intellectual human capital and the birth of U.S.
biotechnology enterprises, American Economic Review, 290-306.
26
The dark line plots the partial derivative of the ln hazard with respect to Radicalness (-2.152 + 0.048(C4)). The shaded line is the 95 percent confidence interval.
Figure 1: Conditional Effect of Technological Radicalness on Mortality
-4
-3
-2
-1
0
1
2
3
10 15 20 25 30 35 40 45 50 55 60 65Concentration Ratio
Log
Haz
ard
Rat
e
27
The dark line plots the partial derivative of the ln hazard with respect to C4 (-0.027 + 0.048(Radicalness)). The shaded line is the 95 percent confidence interval.
Figure2: Conditional Effect of Concentration on Mortality
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
0 5 10 15 20 25 30 35Technological Radicalness
Log
Haz
ard
Rat
e
28
Table 1: Descriptive Statistics
Variable
Mean
S.D.
Failure
0.031
0.174
Age 4.976 3.668 Biotechnology
0.338
0.473
Chemistry
0.367
0.482
Electronics
0.420
0.494
General Purpose
0.198
0.399
Ln (Number of Firms in Industry)
7.782
1.019
Ln (Venture Capital in Industry)
11.357
1.604
Entrepreneur is Inventor
0.673
0.470
Prior Knowledge of Problem Solved
0.880
0.325
Inventor is Tenured at MIT
0.712
0.453
Product Invention
0.664
0.473
Exclusive License
0.817
0.313
Startup Experience on Founding Team
1.018
2.214
Market Experience on Founding Team
12.234
36.850
Technological Radicalness
2.120
1.513
Patent Scope
1.438
0.943
Industry Concentration (C4)
38.100
9.099
N= 128 firms, 834 firm-year spells.
29
Table 2: Weibull Model of Failure Rates for MIT Startups
Variable
(1)
(2)
(3)
(4)
Weibull parameter
0.791**
0.817**
0.885**
0.831**
(0.150) (0.149) (0.155) (0.151) Biotechnology Company
-0.196
0.173
-0.169
0.152
(1.159) (1.217) (0.968) (1.368) Chemistry
0.644
0.561
0.351
1.535
(1.472) (1.478) (1.446) (1.650) Electronics
1.358
1.569*
1.619*
2.537**
(0.872) (0.876) (0.935) (0.992) General Purpose
-0.774
-0.628
-1.004
-0.524
(0.895) (0.867) (0.898) (0.862) Ln (Number of Firms in Industry)
0.306
0.465
0.341
0.621
(0.482) (0.477) (0.464) (0.485) 0.460**
0.412**
0.572**
0.682**
Ln (Venture Capital in Industry)
(0.199) (0.200) (0.206) (0.229) Entrepreneur is Inventor
-1.855**
-1.805**
-1.962**
-1.911**
(0.570) (0.577) (0.572) (0.570)
-0.961
-0.879 -1.434**
-0.491
Prior Knowledge of Problem Solved
(0.632) (0.635) (0.660) (0.680) Inventor is Tenured at MIT
0.332
0.248
0.374
0.489
(0.613) (0.623) (0.606) (0.621) Product Invention
-0.829*
-0.923*
-0.593
-1.200**
(0.480) (0.491) (0.504) (0.548) Exclusive License
0.622
0.790
0.489
0.589
(0.755) (0.769) (0.767) (0.749)
0.145
0.194
0.243
0.279* Startup Experience on Founding Team
(0.146) (0.152) (0.155) (0.160)
-0.043
-0.046 -0.065*
-0.055*
Market Experience on Founding Team
(0.029) (0.029) (0.034) (0.031)
0.061*
0.063*
-0.027
0.048 Industry Concentration (C4)
(0.033) (0.033) (0.049) (0.032) Technological Radicalness
-0.146
-2.152**
(0.194) (0.860) Patent Scope
-0.446
-1.619**
(0.309) (0.621) Concentration * Radicalness
0.048**
30
(0.020) Concentration * Patent Scope
2.877**
(1.236) Log-Likelihood
-50.42
-49.69
-47.767
-46.89
LR Chi-Square
49.27**
50.74**
54.59**
56.34**
Two-sided t-tests: * p<.10 ** p<.05
31
Table 3: Alternative A: Dominant Design
Variable
(1)
(2)
(3)
0.081*
-0.005
0.0001
Industry Concentration (C4)
(0.044) (0.057) (0.062) Technological Radicalness
-2.047*
-2.177**
-2.350**
(1.158) (0.874) (1.128) Concentration * Radicalness
0.047**
0.044*
(0.020) (0.024) Dominant Design
-1.184
-0.587
-0.683
(0.746) (0.712) (0.813) Dominant Design * Radicalness
0.409
0.066
(0.249) (0.273) Log-Likelihood
-48.56
-47.43
-47.39
LR Chi-Square
53.00**
55.26**
55.34**
We control for all of the same variables as are included in Table 2. For ease of exposition, however, we report only the key variables discussed in the focal alternative. The t-tests tests are two-sided: * p<. 10 ** p<.05
32
Table 4: Alternative B: Industry R&D Intensity
Variable
(1)
(2)
(3)
0.073*
0.074*
-0.009
Industry Concentration (C4)
(0.038) (0.038) (0.051) Technological Radicalness
-0.101
-0.220
-2.159**
(0.207) (0.387) (0.817) Concentration * Radicalness
0.049**
(0.019) Industry R&D Intensity
-3.926
-8.119
-7.051
(6.319) (13.109) (12.094) R&D Intensity * Radicalness
2.032
0.023
(5.485) (5.279) Log-Likelihood
-50.23
-50.16
-47.18
LR Chi-Square 49.66**
49.80**
55.76**
We control for all of the same variables as are included in Table 2. For ease of exposition, however, we report only the key variables discussed in the focal alternative. The t-tests tests are two-sided: * p<. 10 ** p<.05.
33
Table 5: Alternative C: Market Growth Rate at Founding
Variable
(1)
(2)
(3)
0.060*
0.061*
-0.023
Industry Concentration (C4)
(0.035) (0.034) (0.050) Technological Radicalness
-0.144
-0.072
-2.101**
(0.195) (0.215) (0.885) Concentration * Radicalness
0.047**
(0.020) Market Growth Rate
0.185
2.182
0.689
(1.781) (3.163) (3.173) Market Growth * Radicalness
-1.424
-0.854
(1.863) (1.941) Log-Likelihood
-50.42
-50.14
-47.64
LR Chi-Square
49.28** 49.84**
54.85**
We control for all of the same variables as are included in Table 2. For ease of exposition, however, we report only the key variables discussed in the focal alternative. The t-tests tests are two-sided: * p<. 10 ** p<.05.
34
Table 6: Alternative D: Technological Maturity
Variable
(1)
(2)
(3)
0.062*
0.054
-0.017
Industry Concentration (C4)
(0.033) (0.034) (0.047) Technological Exploration
-0.163
-4.937**
-5.904**
(0.194) (1.946) (1.981) Concentration * Radicalness
0.043**
(0.021) Age of Patent Class
-0.666*
-2.511**
-2.394**
(0.396) (0.805) (0.836) Age of Patent Class * Radicalness
1.127**
0.939**
(0.457) (0.456) Log-Likelihood
-49.24
-46.01
-43.95
LR Chi-Square
51.64**
58.10**
62.20**
We control for all of the same variables as are included in Table 2. For ease of exposition, however, we report only the key variables discussed in the focal alternative. The t-tests tests are two-sided: * p<. 10 ** p<.05.